Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

M4+ Chip Generation - Speculation Megathread [MERGED]

thenewperson said:
Damn, I didn’t know there was speculation that it could offer 825GB/s, would be wild if the case and yeah I can’t imagine NVIDIA having such an advantage and not making it front and centre in their announcement.

However based on this and this it’s looking like 273GB/s 😐
Click to expand...
If that's accurate then as I wrote above:

crazy dave said:
The lowest it could be is 256-bit bus using 4x32GB modules. That would put it around M4 Pro/Strix Halo bandwidth (~270-300GB/s depending).
Click to expand...

So, indeed they're going with a 256bit bus. With 4x32GB modules, that's the smallest bandwidth they could get away with and still supply 128GB of RAM.

theorist9 said:
Also: It should be noted that NVIDIA says Project Digits will start at $3,000, and include up to a 4 TB SSD.

So if we assume its starting config has a 1 TB SSD: A comparably-spec'd (128 GB RAM/1 TB SSD) 16" M4 Max MBP is $5,000, giving an estimated $4,000 for a hypothetical 1 TB/128 GB M4 Max Studio.

Of course, those really aren't comparable devices—Project Digits will have far more GPU power than an M4 Max, and the M4 Max may have more CPU power. Not sure how their memory bandwidths compare. The M4 Max has 546 GB/s.

Here's an estimate that Digits will have 825 GB/s:

"From the renders shown to the press prior to the Monday night CES keynote at which Nvidia announced the box, the system appeared to feature six LPDDR5x modules. Assuming memory speeds of 8,800 MT/s we'd be looking at around 825GB/s of bandwidth."
https://www.theregister.com/2025/01/07/nvidia_project_digits_mini_pc/"

I've seen competing claims saying it will be much lower. But 825 GB/s seems strong so, if that's correct, why didn't NVIDIA include it in their announcement along with all the other specs?

I wonder if the Project Digits machine will be of interest to gamers, or if the type of GPU performance it offers won't fully translate to gaming. E.g., you will probably be able to buy or build a $3k gaming PC with a 5080 (MSRP $1k, street price TBD), and that will offer 960 GB/s memory bandwidth. [The 5090's memory bandwidth is 1,792 GB/s!]
Click to expand...

My guess is the GB10's FP32 TFLOPS will also be fairly small. Nvidia's workstation graphics have a much higher TOPS to FLOPS ratio and if it is more similar to that (likely), then the GB10 will be closer to base M4 or M4 Pro in terms of graphics performance so no one will be buying it in the hopes of gaming or even doing non-AI compute.

theorist9 said:
From my recollection of comparing an equally-spec'd M2 Max Studio & 16" M2 Max MBP, the MBP was about $1,000 more (consistent with @crazy dave's recollection).

A 16" M4 Max MBP (16‑core CPU, 40‑core GPU) with 128 GB RAM, 4 TB SSD, and a glossy display is $6,000, which implies ≈$5,000 for an equally-spec'd M4 Max Studio.

Though, as has been disucussed, the Studio they announce this year may not be exactly an M4 Max....
Click to expand...

I think it's likely that there will be an M4 Max Studio, it's just we don't know if there will be an M4 Ultra based on the M4 Max or something else. But yes, given what @thenewperson posted, I'd actually say that Apple still has a good shot at offering a better deal here ... maybe. The Studio M4 Max may not compete in Sparse FP4, but will be better in almost every other respect and equal to two DIGITS in bandwidth for twice the price. Depending on what the M4 Ultra is and what it's priced at ... it could be even better.
 
Last edited:
thenewperson said:
Damn, I didn’t know there was speculation that it could offer 825GB/s, would be wild if the case and yeah I can’t imagine NVIDIA having such an advantage and not making it front and centre in their announcement.

However based on this and this it’s looking like 273GB/s 😐
Click to expand...
Yeah, I saw that reddit post; that's what I was thinking of when I wrote "I've seen competing claims saying it will be much lower." I initially linked it in my post, but then deleted it since I didn't know the guy and didn't have time to analyze his argument.

Granted, I also didn't bother to do a close analysis of The Register's claim (thus missing that 128/6 ∉ ℤ, as @crazy dave pointed out), but assumed (apparently incorrectly) that, since they're a tech-focused site that has done some sophsiticated reporting in the past, they were a credible source.
 
Last edited:
Quick question: is it harder to do a memory latency test on macOS? I find it odd that Chips and Cheese only has memory latency and memory bandwidth data from M1.

Or, could it be that people who have access to other Apple hardware haven't run the test yet?
 
Xiao_Xi said:
Quick question: is it harder to do a memory latency test on macOS? I find it odd that Chips and Cheese only has memory latency and memory bandwidth data from M1.

Or, could it be that people who have access to other Apple hardware haven't run the test yet?
Click to expand...
Geekerwan have done cache and memory latency in their newer Apple Silicon reviews.

Screenshot 2025-02-10 at 11.43.16.png
 
  • Like
Reactions: picpicmac, crazy dave, UpsideDownEclair and 5 others
I was actually VERY concerned last night updating to 15.3.1.
My new M4 Mac Mini Pro seemed to be in a boot loop, rebooting in the same part of the progress indicator.
At least two, or three times.
I left it alone, and it finally completed.

Now for a back up solution, and a recovery plan...
 
  • Like
Reactions: norbinhouston and ekwipt
BTW, Apple also uses Chinook cores (which I think are OoO, though I can't put my hands on anything that supports that at the moment), and even some tiny ARM cores like M0/R0, which are NOT OoO.

It is likely that you will never again see a user-facing core that's in-order, though they may be in use for many years to come internally.
 
  • Like
Reactions: Xiao_Xi
Leistung:Energie SPECint 2017 2.png

Thanks to Andrei Frumusanu we have this picture.
At the bottom you have the Arm A55 low power cores.
Down to even 0.18W only.
And you have an A15 E core at 0.44W, but with nearly 4 times the performance it only consumes 60% of the ENERGY.

That's the difference between low power and high efficiency.
So repeat after me: Power usages is not energy usage! 🤓
 
  • Like
Reactions: krell100, crazy dave, Antony Newman and 1 other person
very strange mac studio with M4 Max and M3 ultra that has double everything besides bandwidth
 
Apple: Hey! Let’s run it back 😄

  • Apple’s custom-built UltraFusionpackaging technology uses an embedded silicon interposer that connects two M3 Max dies across more than 10,000 signals, providing over 2.5TB/s of low-latency interprocessor bandwidth, and making M3 Ultra appear as a single chip to software.
 
Last edited:
MayaUser said:
very strange mac studio with M4 Max and M3 ultra that has double everything besides bandwidth
Click to expand...
Yeah, I thought the consensus was that M3 Max didn’t have UltraFusion but apparently it did. And I guess that M4 Max doesn’t then, but that aligns with the rumors that Apple have something extra in store for the top of the line M series for Mac Pro. I guess an announcement for that at WWDC and shipping by end of 2025. Also, they might have an urgent need for the M3 Ultra for their AI server parks so maybe not so surprising to see that after all.
 
Macintosh IIcx said:
Yeah, I thought the consensus was that M3 Max didn’t have UltraFusion but apparently it did.
Click to expand...

It is entirely possible that the Ultra uses a new iteration of the M3 Max chip which includes UltraFusion. Given that Apple did not intent to use the first batch of M3 Max chips for the Ultra, it only makes sense to save the money and chop off the redundant controller part in production.
 
  • Like
Reactions: crazy dave, Macintosh IIcx and tenthousandthings
leman said:
It is entirely possible that the Ultra uses a new iteration of the M3 Max chip which includes UltraFusion. Given that Apple did not intent to use the first batch of M3 Max chips for the Ultra, it only makes sense to save the money and chop off the redundant controller part in production.
Click to expand...
It also "magically" support Thunderbolt 5, so something is definitely new.

They also expressly call out support for USB4 Gen 2 (up to 120Gbps) on both the M4 Max and M3 Ultra Mac Studio.
 
  • Like
Reactions: Antony Newman, crazy dave and Macintosh IIcx
leman said:
It is entirely possible that the Ultra uses a new iteration of the M3 Max chip which includes UltraFusion. Given that Apple did not intent to use the first batch of M3 Max chips for the Ultra, it only makes sense to save the money and chop off the redundant controller part in production.
Click to expand...
Not only that, but maximum addressable memory have dramatically increased. If throw these specs to us without giving out the name, nearly everyone here would have concluded it must have been M4 Max stitched.
 
leman said:
It is entirely possible that the Ultra uses a new iteration of the M3 Max chip which includes UltraFusion. Given that Apple did not intent to use the first batch of M3 Max chips for the Ultra, it only makes sense to save the money and chop off the redundant controller part in production.
Click to expand...
Are you suggesting that the M3 Max for the Ultra is a separate tape out, if ever so slightly different at one side of that chip? If so, I wonder why they didn’t just use M4 Max, but we of course know that the M3 is based on a slightly more advanced node, maybe they needed that for UltraFusion.
 
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back